It matters to be in good hands: the relationship between good governance and pandemic spread inferred from cross-country COVID-19 data

The paper investigates whether and to what extent the quality of governance influences a country’s capability to control the spread of COVID-19. Motivated by the view that governance—encompassing the effectiveness of public services and institutions, policy formulation and implementation, and the credibility of government commitments—can shape compliance and the success of public health interventions, the study posits that better-governed countries will experience lower pandemic spread. Drawing on epidemiological intuition (e.g., reducing the reproduction number via policies that affect contact and recovery rates), the authors hypothesize an inverse relationship between governance quality and measures of COVID-19 spread. The study’s purpose is to provide systematic cross-country empirical evidence, using a large sample and controlling for geographic, demographic, and socio-economic factors, during the early pandemic window of April–September 2020.

Two broad strands of social science research on COVID-19 are highlighted: (1) assessments of the efficacy of specific policy responses (e.g., lockdowns, social distancing) from socioeconomic and behavioral perspectives; and (2) analyses of political regimes and governance as explanatory factors for cross-country variation in pandemic outcomes. Prior findings are mixed: some suggest regime type influences policy efficacy; others find democratic norms matter or that government effectiveness is negatively associated with fatalities. Historical and public health literature emphasizes governance’s role in epidemic control (e.g., HIV studies linking poor governance to higher prevalence). The paper engages with contradictory findings (e.g., claims that less effective governments sometimes responded more decisively) and positions its contribution as testing governance–spread relationships worldwide using infection spread metrics (positive and growth rates) rather than fatalities.

Study design: Cross-sectional econometric analysis using OLS with robust standard errors, estimated separately for each month from April to September 2020 and for the six-month average of dependent variables. Continent fixed effects included. Sub-sample and alternative-specification robustness checks performed. Data and variables: - Sample: 185 countries across six continents (sample size varies by model due to data availability). - Governance measures: World Bank Worldwide Governance Indicators (WGI), 2018 values. Main explanatory variable: Government effectiveness (broadest governance proxy). Additional constituents used in robustness: rule of law, regulatory quality, control of corruption, voice and accountability, political stability. - Dependent variables (April–September 2020): - COVID-19 positive rate = (Total positive cases / Total tests conducted) × 100 (WHO indicates <5% suggests control). - COVID-19 growth rate = ((Ct − C_{t−1}) / C_{t−1}) × 100, where Ct is total confirmed cases per million population. - Controls: - Geographic: ruggedness index, landlocked (dummy), distance from centroid to nearest coast/river, absolute latitude. - Demographic and socio-economic: level of air pollution (PM2.5), urban population percentage, polity ranking (democracy–autocracy score), median age, GDP per capita (PPP, log). Continent fixed effects included in all expanded models. Models: - Positive rate models: (1) Base: positive_rate_i = β0 + β1·govt_effectiveness_i + continent_FE + ε_i. (2) Geographic controls added: + ruggedness, landlock, distance to coast/river, latitude. (3) Full model adds demographic/socio-economic: + urban population, polity ranking, median age, GDP per capita (log), air pollution. - Growth rate models mirror (1)–(3) with growth rate as dependent variable. Estimation and checks: - OLS with robust SEs; models estimated monthly (Apr–Sep 2020) and using six-month averages. - High inter-correlations among WGI constituents documented; therefore, they are not included simultaneously to avoid multicollinearity. - Robustness checks: - Sub-sample analyses: regressions excluding, in turn, Africa, Asia, Europe, North America, Oceania, South America, and OECD from the full sample; and separate regressions for Africa, Asia, Europe, OECD. - Alternative governance measures: repeating the analysis with each WGI constituent separately as the main explanatory variable. - Sensitivity to controls: e.g., re-estimating sub-sample models excluding GDP per capita (log) to assess multicollinearity effects. Assumptions and notes: - Cross-sectional snapshot across months; error terms assumed normal with zero mean and constant variance. Explanatory variables are lagged (2018 governance), mitigating reverse causality concerns.

- Government effectiveness is significantly and negatively associated with COVID-19 spread: - Positive rate models: Significant negative coefficients in five of six months (Apr–Sep 2020); highly significant using the 6‑month average dependent variable. - Growth rate models: Significant negative coefficients in four of six months; highly significant for the 6‑month average. - Magnitudes (standardized effects using 6‑month averages): - Beta coefficient for government effectiveness and positive rate ≈ −0.45 (one SD higher effectiveness → 0.45 SD lower positive rate). - Beta coefficient for government effectiveness and growth rate ≈ −0.85 (one SD higher effectiveness → 0.85 SD lower growth rate). - Control variables: - Median age shows a generally negative association with positive and growth rates when significant, consistent with higher mobility among younger populations increasing spread. - Air pollution shows some influence but with inconsistent signs across months. - Temporal pattern: - The negative association with the growth rate diminishes in later months (some insignificance), plausibly due to compliance fatigue, flattening infection curves, and other unobserved factors; the negative link for the positive rate remains more persistent and strengthens over time. - Robustness: - Sub-sample analyses generally confirm negative coefficients; with all controls, significance is strongest in Africa. After excluding GDP per capita (log), significance improves in most sub-samples (except Asia), suggesting multicollinearity with income may attenuate estimates. - Alternative WGI constituents as main explanatory variables yield consistently negative relationships with positive rates across months: rule of law, regulatory quality, and control of corruption are frequently significant; voice and accountability and political stability show weaker but still negative associations. - Visual evidence (violin and box-density plots) shows lower medians and tighter distributions of positive and growth rates among countries at/above average governance effectiveness relative to those below average. - Overall, results indicate a robust inverse governance–spread relationship that persists after controlling for continent effects and numerous geographic, demographic, and socio-economic factors.

Findings strongly support the hypothesis that better governance—particularly government effectiveness—helps reduce the spread of COVID-19. The likely mechanisms include enhanced policy formulation and implementation capacity, higher credibility of government commitments, and greater public trust leading to higher compliance with mobility-restricting public health measures. The persistent negative association with the positive rate implies sustained advantages in testing strategies and case detection dynamics in better-governed countries. The attenuation of significance for growth-rate models in later months may reflect compliance fatigue, convergence/flattening of epidemic curves near the end of the first wave, or shifting influences of other socio-economic or geographic factors. The consistency of negative coefficients across alternative governance proxies reinforces the systemic role of institutional quality. While the study does not claim causality, the results align with theoretical and historical perspectives emphasizing governance in epidemic control and help reconcile mixed prior findings by highlighting that trust-based, effective governance may outperform heavy-handed responsiveness in sustaining control over time.

The study provides cross-country empirical evidence (185 countries; Apr–Sep 2020) that higher governance quality—measured primarily by government effectiveness and corroborated by other WGI dimensions—is associated with lower COVID-19 spread, as measured by positive and growth rates. The relationship is robust to continent fixed effects and a wide set of geographic, demographic, and socio-economic controls, and it holds in various sub-samples and under alternative governance measures. These results substantiate the notion that being “in good hands” matters for pandemic control. Future research directions include: employing panel-data or dynamic causal frameworks to address temporal evolution and causality; exploring mechanisms linking governance to compliance, trust, and mobility; testing alternative democracy measures (e.g., V‑Dem) and institutional dimensions; and integrating additional epidemiological and policy-response variables to disentangle pathways through which governance impacts pandemic control.

- Design: Cross-sectional OLS snapshots across months cannot establish causality; results indicate associations only. - Data timing: Governance indicators are from 2018; although lagging mitigates reverse causality, it may not capture governance changes by 2020. - Multicollinearity: High inter-correlations among WGI dimensions and with income (GDP per capita) can attenuate significance; addressed by not including all WGI together and by sensitivity checks excluding GDP per capita. - Measurement and data quality: Variation in testing practices affects the positive rate; reporting differences across countries can influence both dependent and control variables. Sample sizes vary by model due to data availability. - Temporal dynamics: Later-month attenuation in growth-rate significance suggests unobserved time-varying influences (e.g., compliance fatigue, policy changes) that cross-sectional models cannot capture. Panel methods were deemed impractical given data constraints. - Control specification: Alternative measures (e.g., democracy indices such as V‑Dem) can alter precision due to correlations with governance; results are qualitatively robust but somewhat attenuated with alternative democracy measures.